A self-supervised speech representation model from Facebook, pre-trained on 960 hours of LibriSpeech for audio understanding.